[{"data":1,"prerenderedAt":117},["ShallowReactive",2],{"\u002Fblog\u002Fai-agent-governance-scenarios-part-4-the-efficiency-shortcut":3},{"id":4,"title":5,"author":6,"body":7,"category":99,"date":100,"description":101,"draft":102,"excerpt":103,"extension":104,"heroImage":103,"heroImageAlt":103,"meta":105,"navigation":106,"path":107,"seo":108,"slug":103,"stem":109,"tags":110,"__hash__":116},"blog\u002Fblog\u002Fai-agent-governance-scenarios-part-4-the-efficiency-shortcut.md","AI Agent Governance Scenarios, Part 4: The \"Efficiency Shortcut\"","QikSolve",{"type":8,"value":9,"toc":91},"minimark",[10,14,19,22,34,38,45,51,57,63,67,70,74],[11,12,13],"p",{},"This is the fourth in a five-part series of practical AI-governance scenarios for quality\nprofessionals.",[15,16,18],"h2",{"id":17},"the-scenario","The scenario",[11,20,21],{},"An organisation implemented an AI agent to assist QA reviewers with batch record review. The\nintended workflow was structured and sequential: the agent flags potential issues, the reviewer\nevaluates those flags, and the reviewer makes the final determination — the agent accelerates the\nreview; it does not replace it.",[11,23,24,25,29,30,33],{},"During a routine review meeting, quality leadership discovers that a number of reviewers have\nquietly changed how they use the agent. Instead of running the structured workflow, they ask a\nsingle direct question: ",[26,27,28],"em",{},"\"Are there any compliance concerns in this batch record?\""," The agent\nreturns a short, reassuring summary — ",[26,31,32],{},"\"No major documentation issues detected\""," — and the reviewer\nproceeds with minimal further evaluation. This approach has become common practice. The system\nconfiguration has not changed. No procedure has been updated. No change control has been raised. Yet\nthe effective process has shifted fundamentally.",[15,35,37],{"id":36},"is-this-acceptable-and-what-would-you-do-about-it","Is this acceptable, and what would you do about it?",[11,39,40,44],{},[41,42,43],"strong",{},"Issue 1 — the defined review workflow is bypassed."," The agent was validated to support a\nstructured process; that structure defines the scope within which its outputs can be considered\nreliable. A general question invokes a surface-level interpretation rather than the systematic\ncheck the workflow was designed to run, and the reviewer has no visibility into what was actually\nevaluated. From a GMP perspective this is a process deviation, whether or not anyone documented it as\none — the gap between written procedure and actual practice being invisible to the quality system\nmakes it worse, not better.",[11,46,47,50],{},[41,48,49],{},"Issue 2 — the agent is used as a decision shortcut."," When a reassuring summary is accepted with\nminimal further review, the agent has effectively become a decision authority rather than a review\nassistant — a fundamental inversion of the intended relationship. \"No major issues detected\" is not\na QA release decision; it is an output that requires human interpretation to be meaningful, and the\ndistinction between \"the AI found no issues\" and \"the record is acceptable\" is being collapsed.",[11,52,53,56],{},[41,54,55],{},"Issue 3 — human verification is reduced."," GMP requires meaningful, independent verification, not\ncursory confirmation of what an agent has already reported. When a reviewer's evaluation is shaped\nprimarily by the agent's conclusion, the independence of that verification is compromised — and a\nreviewer expecting a clean record may apply less scrutiny than one approaching the record fresh, a\nwell-documented cognitive effect amplified by algorithmic authority.",[11,58,59,62],{},[41,60,61],{},"Issue 4 — informal behaviour change without governance."," This is arguably the most significant\nissue: the written procedure describes the approved workflow, but actual practice has diverged, and\nthat gap is invisible without active oversight. Behavioural changes to how a tool is used constitute\nan effective process change, even without a formal update — governance has to extend beyond initial\nvalidation to include periodic review of interaction logs, competency checks, and clear escalation\npathways for identified workarounds.",[15,64,66],{"id":65},"two-different-failure-modes","Two different failure modes",[11,68,69],{},"This scenario and the \"Game Changer Agent\" scenario illustrate two distinct failure types. A\ntechnology-governance failure is a change in how the system itself is controlled — visible, and\nusually caught by existing change control. A human behavioural-drift failure — this one — is a\nchange in how people interact with a system that has not technically changed at all: no alarm\ntriggers, no change control initiates, and the quality system may be entirely unaware anything has\nshifted. AI risk in GMP is primarily about governance and behaviour, not the model itself, and the\nmost significant risks often come from how people choose to use a tool rather than from what the\ntool does.",[15,71,73],{"id":72},"related-reading","Related reading",[75,76,77,85],"ul",{},[78,79,80],"li",{},[81,82,84],"a",{"href":83},"\u002Fblog\u002Fai-agent-governance-scenarios-part-3-the-smart-sop-assistant","Part 3: The \"Smart SOP Assistant\"",[78,86,87],{},[81,88,90],{"href":89},"\u002Fblog\u002Fai-agent-governance-scenarios-part-5-the-inspection-question","Part 5: The Inspection Question",{"title":92,"searchDepth":93,"depth":93,"links":94},"",2,[95,96,97,98],{"id":17,"depth":93,"text":18},{"id":36,"depth":93,"text":37},{"id":65,"depth":93,"text":66},{"id":72,"depth":93,"text":73},"AI Governance","2026-09-10","The system did not change and no change control was raised, but the way people actually used it drifted from a structured review workflow to a single question.",false,null,"md",{},true,"\u002Fblog\u002Fai-agent-governance-scenarios-part-4-the-efficiency-shortcut",{"title":5,"description":101},"blog\u002Fai-agent-governance-scenarios-part-4-the-efficiency-shortcut",[111,112,113,114,115],"ai-agents","gxp","workshop-scenario","human-oversight","behavioural-drift","JV2o4JfeDE1lfKM62xXRCfAbm80lvWSj4_d4Puq8ZNQ",1789037363494]